summaryrefslogtreecommitdiffstats
path: root/arch/blackfin/mm
diff options
context:
space:
mode:
Diffstat (limited to 'arch/blackfin/mm')
-rw-r--r--arch/blackfin/mm/init.c61
-rw-r--r--arch/blackfin/mm/isram-driver.c101
-rw-r--r--arch/blackfin/mm/sram-alloc.c31
3 files changed, 85 insertions, 108 deletions
diff --git a/arch/blackfin/mm/init.c b/arch/blackfin/mm/init.c
index bb9c98f9cb5..bb4e8fff4b5 100644
--- a/arch/blackfin/mm/init.c
+++ b/arch/blackfin/mm/init.c
@@ -4,6 +4,7 @@
* Licensed under the GPL-2 or later.
*/
+#include <linux/gfp.h>
#include <linux/swap.h>
#include <linux/bootmem.h>
#include <linux/uaccess.h>
@@ -14,23 +15,11 @@
#include "blackfin_sram.h"
/*
- * BAD_PAGE is the page that is used for page faults when linux
- * is out-of-memory. Older versions of linux just did a
- * do_exit(), but using this instead means there is less risk
- * for a process dying in kernel mode, possibly leaving a inode
- * unused etc..
- *
- * BAD_PAGETABLE is the accompanying page-table: it is initialized
- * to point to BAD_PAGE entries.
- *
- * ZERO_PAGE is a special page that is used for zero-initialized
- * data and COW.
+ * ZERO_PAGE is a special page that is used for zero-initialized data and COW.
+ * Let the bss do its zero-init magic so we don't have to do it ourselves.
*/
-static unsigned long empty_bad_page_table;
-
-static unsigned long empty_bad_page;
-
-static unsigned long empty_zero_page;
+char empty_zero_page[PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
+EXPORT_SYMBOL(empty_zero_page);
#ifndef CONFIG_EXCEPTION_L1_SCRATCH
#if defined CONFIG_SYSCALL_TAB_L1
@@ -51,40 +40,26 @@ EXPORT_SYMBOL(cpu_pda);
void __init paging_init(void)
{
/*
- * make sure start_mem is page aligned, otherwise bootmem and
- * page_alloc get different views og the world
+ * make sure start_mem is page aligned, otherwise bootmem and
+ * page_alloc get different views of the world
*/
unsigned long end_mem = memory_end & PAGE_MASK;
- pr_debug("start_mem is %#lx virtual_end is %#lx\n", PAGE_ALIGN(memory_start), end_mem);
-
- /*
- * initialize the bad page table and bad page to point
- * to a couple of allocated pages
- */
- empty_bad_page_table = (unsigned long)alloc_bootmem_pages(PAGE_SIZE);
- empty_bad_page = (unsigned long)alloc_bootmem_pages(PAGE_SIZE);
- empty_zero_page = (unsigned long)alloc_bootmem_pages(PAGE_SIZE);
- memset((void *)empty_zero_page, 0, PAGE_SIZE);
+ unsigned long zones_size[MAX_NR_ZONES] = {
+ [0] = 0,
+ [ZONE_DMA] = (end_mem - PAGE_OFFSET) >> PAGE_SHIFT,
+ [ZONE_NORMAL] = 0,
+#ifdef CONFIG_HIGHMEM
+ [ZONE_HIGHMEM] = 0,
+#endif
+ };
- /*
- * Set up SFC/DFC registers (user data space)
- */
+ /* Set up SFC/DFC registers (user data space) */
set_fs(KERNEL_DS);
- pr_debug("free_area_init -> start_mem is %#lx virtual_end is %#lx\n",
+ pr_debug("free_area_init -> start_mem is %#lx virtual_end is %#lx\n",
PAGE_ALIGN(memory_start), end_mem);
-
- {
- unsigned long zones_size[MAX_NR_ZONES] = { 0, };
-
- zones_size[ZONE_DMA] = (end_mem - PAGE_OFFSET) >> PAGE_SHIFT;
- zones_size[ZONE_NORMAL] = 0;
-#ifdef CONFIG_HIGHMEM
- zones_size[ZONE_HIGHMEM] = 0;
-#endif
- free_area_init(zones_size);
- }
+ free_area_init(zones_size);
}
asmlinkage void __init init_pda(void)
diff --git a/arch/blackfin/mm/isram-driver.c b/arch/blackfin/mm/isram-driver.c
index 84cdc5a1c13..7e2e674ed44 100644
--- a/arch/blackfin/mm/isram-driver.c
+++ b/arch/blackfin/mm/isram-driver.c
@@ -11,6 +11,7 @@
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/types.h>
+#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/sched.h>
@@ -42,13 +43,12 @@ static DEFINE_SPINLOCK(dtest_lock);
/* Takes a void pointer */
#define IADDR2DTEST(x) \
({ unsigned long __addr = (unsigned long)(x); \
- (__addr & 0x47F8) | /* address bits 14 & 10:3 */ \
- (__addr & 0x8000) << 23 | /* Bank A/B */ \
- (__addr & 0x0800) << 15 | /* address bit 11 */ \
- (__addr & 0x3000) << 4 | /* address bits 13:12 */ \
- (__addr & 0x8000) << 8 | /* address bit 15 */ \
- (0x1000000) | /* instruction access = 1 */ \
- (0x4); /* data array = 1 */ \
+ ((__addr & (1 << 11)) << (26 - 11)) | /* addr bit 11 (Way0/Way1) */ \
+ (1 << 24) | /* instruction access = 1 */ \
+ ((__addr & (1 << 15)) << (23 - 15)) | /* addr bit 15 (Data Bank) */ \
+ ((__addr & (3 << 12)) << (16 - 12)) | /* addr bits 13:12 (Subbank) */ \
+ (__addr & 0x47F8) | /* addr bits 14 & 10:3 */ \
+ (1 << 2); /* data array = 1 */ \
})
/* Takes a pointer, and returns the offset (in bits) which things should be shifted */
@@ -62,7 +62,7 @@ static void isram_write(const void *addr, uint64_t data)
uint32_t cmd;
unsigned long flags;
- if (addr >= (void *)(L1_CODE_START + L1_CODE_LENGTH))
+ if (unlikely(addr >= (void *)(L1_CODE_START + L1_CODE_LENGTH)))
return;
cmd = IADDR2DTEST(addr) | 2; /* write */
@@ -93,7 +93,7 @@ static uint64_t isram_read(const void *addr)
unsigned long flags;
uint64_t ret;
- if (addr > (void *)(L1_CODE_START + L1_CODE_LENGTH))
+ if (unlikely(addr > (void *)(L1_CODE_START + L1_CODE_LENGTH)))
return 0;
cmd = IADDR2DTEST(addr) | 0; /* read */
@@ -120,7 +120,7 @@ static bool isram_check_addr(const void *addr, size_t n)
{
if ((addr >= (void *)L1_CODE_START) &&
(addr < (void *)(L1_CODE_START + L1_CODE_LENGTH))) {
- if ((addr + n) > (void *)(L1_CODE_START + L1_CODE_LENGTH)) {
+ if (unlikely((addr + n) > (void *)(L1_CODE_START + L1_CODE_LENGTH))) {
show_stack(NULL, NULL);
pr_err("copy involving %p length (%zu) too long\n", addr, n);
}
@@ -195,7 +195,7 @@ EXPORT_SYMBOL(isram_memcpy);
#ifdef CONFIG_BFIN_ISRAM_SELF_TEST
-#define TEST_LEN 0x100
+static int test_len = 0x20000;
static __init void hex_dump(unsigned char *buf, int len)
{
@@ -211,15 +211,15 @@ static __init int isram_read_test(char *sdram, void *l1inst)
pr_info("INFO: running isram_read tests\n");
/* setup some different data to play with */
- for (i = 0; i < TEST_LEN; ++i)
- sdram[i] = i;
- dma_memcpy(l1inst, sdram, TEST_LEN);
+ for (i = 0; i < test_len; ++i)
+ sdram[i] = i % 255;
+ dma_memcpy(l1inst, sdram, test_len);
/* make sure we can read the L1 inst */
- for (i = 0; i < TEST_LEN; i += sizeof(uint64_t)) {
+ for (i = 0; i < test_len; i += sizeof(uint64_t)) {
data1 = isram_read(l1inst + i);
memcpy(&data2, sdram + i, sizeof(data2));
- if (memcmp(&data1, &data2, sizeof(uint64_t))) {
+ if (data1 != data2) {
pr_err("FAIL: isram_read(%p) returned %#llx but wanted %#llx\n",
l1inst + i, data1, data2);
++ret;
@@ -237,25 +237,25 @@ static __init int isram_write_test(char *sdram, void *l1inst)
pr_info("INFO: running isram_write tests\n");
/* setup some different data to play with */
- memset(sdram, 0, TEST_LEN * 2);
- dma_memcpy(l1inst, sdram, TEST_LEN);
- for (i = 0; i < TEST_LEN; ++i)
- sdram[i] = i;
+ memset(sdram, 0, test_len * 2);
+ dma_memcpy(l1inst, sdram, test_len);
+ for (i = 0; i < test_len; ++i)
+ sdram[i] = i % 255;
/* make sure we can write the L1 inst */
- for (i = 0; i < TEST_LEN; i += sizeof(uint64_t)) {
+ for (i = 0; i < test_len; i += sizeof(uint64_t)) {
memcpy(&data1, sdram + i, sizeof(data1));
isram_write(l1inst + i, data1);
data2 = isram_read(l1inst + i);
- if (memcmp(&data1, &data2, sizeof(uint64_t))) {
+ if (data1 != data2) {
pr_err("FAIL: isram_write(%p, %#llx) != %#llx\n",
l1inst + i, data1, data2);
++ret;
}
}
- dma_memcpy(sdram + TEST_LEN, l1inst, TEST_LEN);
- if (memcmp(sdram, sdram + TEST_LEN, TEST_LEN)) {
+ dma_memcpy(sdram + test_len, l1inst, test_len);
+ if (memcmp(sdram, sdram + test_len, test_len)) {
pr_err("FAIL: isram_write() did not work properly\n");
++ret;
}
@@ -267,12 +267,12 @@ static __init int
_isram_memcpy_test(char pattern, void *sdram, void *l1inst, const char *smemcpy,
void *(*fmemcpy)(void *, const void *, size_t))
{
- memset(sdram, pattern, TEST_LEN);
- fmemcpy(l1inst, sdram, TEST_LEN);
- fmemcpy(sdram + TEST_LEN, l1inst, TEST_LEN);
- if (memcmp(sdram, sdram + TEST_LEN, TEST_LEN)) {
+ memset(sdram, pattern, test_len);
+ fmemcpy(l1inst, sdram, test_len);
+ fmemcpy(sdram + test_len, l1inst, test_len);
+ if (memcmp(sdram, sdram + test_len, test_len)) {
pr_err("FAIL: %s(%p <=> %p, %#x) failed (data is %#x)\n",
- smemcpy, l1inst, sdram, TEST_LEN, pattern);
+ smemcpy, l1inst, sdram, test_len, pattern);
return 1;
}
return 0;
@@ -291,12 +291,13 @@ static __init int isram_memcpy_test(char *sdram, void *l1inst)
/* check read of small, unaligned, and hardware 64bit limits */
pr_info("INFO: running isram_memcpy (read) tests\n");
- for (i = 0; i < TEST_LEN; ++i)
- sdram[i] = i;
- dma_memcpy(l1inst, sdram, TEST_LEN);
+ /* setup some different data to play with */
+ for (i = 0; i < test_len; ++i)
+ sdram[i] = i % 255;
+ dma_memcpy(l1inst, sdram, test_len);
thisret = 0;
- for (i = 0; i < TEST_LEN - 32; ++i) {
+ for (i = 0; i < test_len - 32; ++i) {
unsigned char cmp[32];
for (j = 1; j <= 32; ++j) {
memset(cmp, 0, sizeof(cmp));
@@ -309,7 +310,7 @@ static __init int isram_memcpy_test(char *sdram, void *l1inst)
pr_cont("\n");
if (++thisret > 20) {
pr_err("FAIL: skipping remaining series\n");
- i = TEST_LEN;
+ i = test_len;
break;
}
}
@@ -320,11 +321,11 @@ static __init int isram_memcpy_test(char *sdram, void *l1inst)
/* check write of small, unaligned, and hardware 64bit limits */
pr_info("INFO: running isram_memcpy (write) tests\n");
- memset(sdram + TEST_LEN, 0, TEST_LEN);
- dma_memcpy(l1inst, sdram + TEST_LEN, TEST_LEN);
+ memset(sdram + test_len, 0, test_len);
+ dma_memcpy(l1inst, sdram + test_len, test_len);
thisret = 0;
- for (i = 0; i < TEST_LEN - 32; ++i) {
+ for (i = 0; i < test_len - 32; ++i) {
unsigned char cmp[32];
for (j = 1; j <= 32; ++j) {
isram_memcpy(l1inst + i, sdram + i, j);
@@ -337,7 +338,7 @@ static __init int isram_memcpy_test(char *sdram, void *l1inst)
pr_cont("\n");
if (++thisret > 20) {
pr_err("FAIL: skipping remaining series\n");
- i = TEST_LEN;
+ i = test_len;
break;
}
}
@@ -354,22 +355,30 @@ static __init int isram_test_init(void)
char *sdram;
void *l1inst;
- sdram = kmalloc(TEST_LEN * 2, GFP_KERNEL);
- if (!sdram) {
- pr_warning("SKIP: could not allocate sdram\n");
- return 0;
+ /* Try to test as much of L1SRAM as possible */
+ while (test_len) {
+ test_len >>= 1;
+ l1inst = l1_inst_sram_alloc(test_len);
+ if (l1inst)
+ break;
}
-
- l1inst = l1_inst_sram_alloc(TEST_LEN);
if (!l1inst) {
- kfree(sdram);
pr_warning("SKIP: could not allocate L1 inst\n");
return 0;
}
+ pr_info("INFO: testing %#x bytes (%p - %p)\n",
+ test_len, l1inst, l1inst + test_len);
+
+ sdram = kmalloc(test_len * 2, GFP_KERNEL);
+ if (!sdram) {
+ sram_free(l1inst);
+ pr_warning("SKIP: could not allocate sdram\n");
+ return 0;
+ }
/* sanity check initial L1 inst state */
ret = 1;
- pr_info("INFO: running initial dma_memcpy checks\n");
+ pr_info("INFO: running initial dma_memcpy checks %p\n", sdram);
if (_isram_memcpy_test(0xa, sdram, l1inst, dma_memcpy))
goto abort;
if (_isram_memcpy_test(0x5, sdram, l1inst, dma_memcpy))
diff --git a/arch/blackfin/mm/sram-alloc.c b/arch/blackfin/mm/sram-alloc.c
index f068c11ea98..627e04b5ba9 100644
--- a/arch/blackfin/mm/sram-alloc.c
+++ b/arch/blackfin/mm/sram-alloc.c
@@ -17,6 +17,7 @@
#include <linux/proc_fs.h>
#include <linux/spinlock.h>
#include <linux/rtc.h>
+#include <linux/slab.h>
#include <asm/blackfin.h>
#include <asm/mem_map.h>
#include "blackfin_sram.h"
@@ -255,7 +256,8 @@ static void *_sram_alloc(size_t size, struct sram_piece *pfree_head,
plast->next = pslot->next;
pavail = pslot;
} else {
- pavail = kmem_cache_alloc(sram_piece_cache, GFP_KERNEL);
+ /* use atomic so our L1 allocator can be used atomically */
+ pavail = kmem_cache_alloc(sram_piece_cache, GFP_ATOMIC);
if (!pavail)
return NULL;
@@ -402,7 +404,7 @@ void *l1_data_A_sram_alloc(size_t size)
void *addr;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1_data_sram_lock, cpu), flags);
@@ -411,7 +413,6 @@ void *l1_data_A_sram_alloc(size_t size)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1_data_sram_lock, cpu), flags);
- put_cpu();
pr_debug("Allocated address in l1_data_A_sram_alloc is 0x%lx+0x%lx\n",
(long unsigned int)addr, size);
@@ -430,7 +431,7 @@ int l1_data_A_sram_free(const void *addr)
int ret;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1_data_sram_lock, cpu), flags);
@@ -439,7 +440,6 @@ int l1_data_A_sram_free(const void *addr)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1_data_sram_lock, cpu), flags);
- put_cpu();
return ret;
#else
@@ -455,7 +455,7 @@ void *l1_data_B_sram_alloc(size_t size)
void *addr;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1_data_sram_lock, cpu), flags);
@@ -464,7 +464,6 @@ void *l1_data_B_sram_alloc(size_t size)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1_data_sram_lock, cpu), flags);
- put_cpu();
pr_debug("Allocated address in l1_data_B_sram_alloc is 0x%lx+0x%lx\n",
(long unsigned int)addr, size);
@@ -483,7 +482,7 @@ int l1_data_B_sram_free(const void *addr)
int ret;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1_data_sram_lock, cpu), flags);
@@ -492,7 +491,6 @@ int l1_data_B_sram_free(const void *addr)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1_data_sram_lock, cpu), flags);
- put_cpu();
return ret;
#else
@@ -540,7 +538,7 @@ void *l1_inst_sram_alloc(size_t size)
void *addr;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1_inst_sram_lock, cpu), flags);
@@ -549,7 +547,6 @@ void *l1_inst_sram_alloc(size_t size)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1_inst_sram_lock, cpu), flags);
- put_cpu();
pr_debug("Allocated address in l1_inst_sram_alloc is 0x%lx+0x%lx\n",
(long unsigned int)addr, size);
@@ -568,7 +565,7 @@ int l1_inst_sram_free(const void *addr)
int ret;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1_inst_sram_lock, cpu), flags);
@@ -577,7 +574,6 @@ int l1_inst_sram_free(const void *addr)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1_inst_sram_lock, cpu), flags);
- put_cpu();
return ret;
#else
@@ -593,7 +589,7 @@ void *l1sram_alloc(size_t size)
void *addr;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1sram_lock, cpu), flags);
@@ -602,7 +598,6 @@ void *l1sram_alloc(size_t size)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1sram_lock, cpu), flags);
- put_cpu();
return addr;
}
@@ -614,7 +609,7 @@ void *l1sram_alloc_max(size_t *psize)
void *addr;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1sram_lock, cpu), flags);
@@ -623,7 +618,6 @@ void *l1sram_alloc_max(size_t *psize)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1sram_lock, cpu), flags);
- put_cpu();
return addr;
}
@@ -635,7 +629,7 @@ int l1sram_free(const void *addr)
int ret;
unsigned int cpu;
- cpu = get_cpu();
+ cpu = smp_processor_id();
/* add mutex operation */
spin_lock_irqsave(&per_cpu(l1sram_lock, cpu), flags);
@@ -644,7 +638,6 @@ int l1sram_free(const void *addr)
/* add mutex operation */
spin_unlock_irqrestore(&per_cpu(l1sram_lock, cpu), flags);
- put_cpu();
return ret;
}
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2024-12-03 19:58:35 +0000